Skip to content

Connect an AI client (MCP)

PostMCP exposes 16 publishing tools over the Model Context Protocol. Any MCP client, including Claude, ChatGPT, Cursor and Claude Code, can list your accounts, check the queue and create, schedule or retry posts.

On this page
  1. API key and server URL
  2. Claude
  3. ChatGPT
  4. Claude Desktop and Cursor
  5. Claude Code and terminals
  6. Running the server yourself
  7. What the model can see

API key and server URL

Open AI Integration. Your key is shown at the top with a ready-made server URL:

MCP server URL
https://mcp.postmcpai.com/mcp?apikey=YOUR_API_KEY

The key is bound to the workspace it was issued from, so a client using it acts on that workspace without being told which. To point it at another workspace, append &projectId=WORKSPACE_ID or regenerate the key from inside that workspace.

Claude

  1. 1
    Copy the server URL from AI Integration
  2. 2
    In Claude, open Settings, then Connectors, then Add custom connector
    Paste the URL and save. Claude discovers the tools on its own.
  3. 3
    Test it
    Ask: “Check my connected social media accounts on PostMCP AI and draft a tweet announcing our new product release.”

Full walkthrough: Claude integration guide.

ChatGPT

  1. 1
    Turn on Developer Mode
    ChatGPT Settings, then Advanced, then Developer Mode. Custom connectors need a Plus or Pro plan.
  2. 2
    Create a connector
    Settings, then Connectors, then Create Connector. Paste the server URL and select No Auth, since the key is already in the URL.
  3. 3
    Test it
    Ask: “Fetch my social accounts status from PostMCP AI and draft a LinkedIn post about AI workflow automation.”

Full walkthrough: ChatGPT integration guide.

Claude Desktop and Cursor

Both take an mcpServers block. The bridge package mcp-remote connects a local client to the hosted server. Replace the placeholder with your key, or copy the filled-in version from AI Integration.

claude_desktop_config.json · Cursor MCP settings
{
  "mcpServers": {
    "postmcpai": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.postmcpai.com/mcp?apikey=YOUR_API_KEY"
      ]
    }
  }
}

Where the config file is

  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
  • Windows: %APPDATA%\Claude\claude_desktop_config.json
  • Cursor: Settings, then MCP, then Add new MCP server, and paste the same block. See the Cursor guide.

Restart the app after saving. The PostMCP tools appear in the client’s tool list.

Claude Code and terminals

Terminal
npx postmcp-cli login
claude mcp add postmcpai -- npx -y mcp-remote "https://mcp.postmcpai.com/mcp?apikey=YOUR_API_KEY"

The first command signs your terminal in. The second registers the tools with Claude Code. For agents that use skills instead of MCP, npx skills add postmcpai/skill installs the PostMCP skill.

Running the server yourself

The server is open source and published on npm as @postmcpai/server. Run it locally over stdio with your key in the environment, or set PORT to serve it over streamable HTTP from your own infrastructure.

claude_desktop_config.json · local server
{
  "mcpServers": {
    "postmcpai": {
      "command": "npx",
      "args": ["-y", "@postmcpai/server"],
      "env": {
        "POSTMCPAI_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

Tool-by-tool parameters, responses and error cases are in the API reference. Transports and hosting are on the MCP server page.

What the model can see

Tool names, the arguments it chooses and the results. That is the whole surface: enough to publish a post, not enough to take over an account.

  • Your social OAuth tokens stay encrypted in the PostMCP backend. They are never sent to the MCP server or the model.
  • Your API key lives in the client configuration and authenticates calls to the backend. It is never returned as tool output.
  • Mutating tools such as create_post and delete_post go through the client’s normal approval prompt. Read tools are safe to run unattended.